st: System GMM line

[Francesco Grigoli <francesco.grigoli.82@gmail.com>]

Dear all,

I am experiencing a problem with the System GMM estimation.

If I run a cross correlation analysis I get that income per capita in
PPP terms and public investment as a percentage of GDP are negatively
correlated:


                    | ln_gdppc public~o
-------------+------------------
    ln_gdppc |   1.0000
public_inv~o|  -0.1646   1.0000


If I run a fixed effects panel regression, I get a negative
coefficient on the income per capita variable. However, if I run a
sytem GMM, which should take care of the endogeneity between income
and public investment, i get a positive and significant coefficient on
the income variable. I don't expect it to be significantly negative,
but the positive sign is very weird, given also that with other
methodologies it is negative (and significant).

Here is the line I used:

xi: xtabond2 public_investment_ratio l.public_investment_ratio
quality_of_governance ln_cia_land_area ln_wdi_pop left_party_largest
penn_pi ln_gdppc i.year, gmm(public_investment_ratio penn_pi ln_gdppc,
lag(1 5) collapse) iv(quality_of_governance ln_cia_land_area
ln_wdi_pop left_party_largest i.year) robust


Whichever lag structure I use, I get the positive and significant
coefficient. Am I doing something wrong with the code for the system
GMM?

Best,
Francesco
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/